?methylTFR, and the copyright holder is Saarland University.Changes from the Bioconductor review.
API CHANGES
Exported functions now use camelCase consistently. Arguments are unchanged.
| Old | New |
|---|---|
| read_methylome() | readMethylome() |
| run_methyltfr() | runMethylTFR() |
| run_methylTFR_RnBeads() | runMethylTFRFromRnBeads() |
| differential_deviation_test() | differentialDeviationTest() |
| addGCBintoMethylome() | addGCBinToMethylome() |
NEW FEATURES
runMethylTFR() and runMethylTFRFromRnBeads() take a tempDir
argument. Temporary HDF5 files are written under tempdir() by default,
not in the working directory, and are deleted when the run ends.computeDeviation() takes extendBy (default 130) and windowBreaks
(default c(-250, -200, -25, 25, 200, 250)), which were hard-coded.msites must have score and coverage columns,
gc_dist must have a GC_bin column, and every motif in gcfreqs must
have binding sites in tf_bindsites. Error messages name the function
that produces the missing object.BUG FIXES
NA when the centre bin or either outer flank
bin has no CpG. Before, an empty bin shifted the bins and gave a wrong
ratio.mean_difference in differentialDeviationTest() is
now signed (first group minus second group). It was the absolute value.cleanMem(), which forced garbage collection after every
block written, and the unused computeColZScore().DOCUMENTATION
?methylTFR).CHANGES
Authors@R lists the funders (ERA-NET Transcan-Neu III - EPILUNAR,
grant 01KT2409; Saarland University NanoBioMed Young Investigator Grant)
with the fnd role.inst/CITATION rewritten: the citation reads "Gunduz IB, Murugan SK,
Mueller F (year). methylTFR: ... R package version x.y.z.", takes the
version and year from DESCRIPTION, and carries the Bioconductor DOI.CITATION.cff (excluded from the build via .Rbuildignore).BUG FIXES
rnb_sample_ids() no longer calls RnBeads::samples(), which is not
exported by all RnBeads versions (R CMD check: "Missing or unexported
object"). Sample identifiers are read from the identifiers.column of
RnBeads::pheno(), with the previous fallbacks unchanged.inst/scripts/vignette_data.R call
getGenomeGC() without an argument, which works with every version of
the annotation packages.BUG FIXES
read_methylome(type = "encode") returned methylation scores that were not
fractions. parse_encode() divided column 11 by column 10, but in a bedMethyl
file column 10 is read coverage and column 11 is the percentage of reads
methylated, so a fully methylated CpG covered by five reads scored 20. The
score is now the percentage over 100.parse_encode() skipped the first line of every file unconditionally. Files
downloaded from ENCODE carry no header, so the first record was silently
discarded. The header is now detected.read_methylome() stops if the parsed scores fall outside 0 to 1 rather than
passing them to the deviation calculation.run_methylTFR_RnBeads() example built an RnBiseqSet without setting
identifiers.column, so samples() returned row numbers and the example
stopped on the sample-identifier check. It now sets the option before
building the set and restores it afterwards.encode.tsv.gz example has been rebuilt. Columns 10 and 11 are
now named coverage and percentMeth rather than blockCount and
blockSizes, and the percentages span 0 to 100, so the tests exercise the
case the old example passed by luck.NEW FEATURES
run_methylTFR_RnBeads(), which runs the methylTFR workflow directly
on a preprocessed RnBeads object instead of per-sample BED files. Samples are
read one column at a time, so disk-backed RnBeads sets are never loaded into
memory in full.computeZScoreVariability(), which ranks TF motifs by how much their
activity varies across samples and tests each motif against a chi-squared
null. Deviation scores are first calibrated against a within-sample null
estimated across motifs, so that a variability above 1 is interpretable as
"more variable than background".methylTFRdeviations objects returned by run_methyltfr() and
run_methylTFR_RnBeads() now carry a third assay, expected, holding the
GC-derived expected deviations. These were previously computed and then
discarded.BUG FIXES
run_methyltfr() no longer rejects .csv sample annotation files. The
.tsv branch's else clause caught every .csv file and raised an error
after the file had already been read.differential_deviation_test() now falls back to colnames(deviations)
when groups is NULL, instead of colnames(groups), which was always NULL.
Group labels are also validated against the number of columns.cov_threshold in run_methyltfr() now sets
the documented default of 1 rather than 5.DOCUMENTATION
ComplexHeatmap instead of ggplot2, with columns split by cell type
and rows clustered.INTERNAL
methyltfr_core(), so the file-based and RnBeads-based workflows are
guaranteed to produce identical results for the same methylation calls.NEW FEATURES
NEWS.md file to track changes to the package.